PHENOPSIS, an automated platform for reproducible phenotyping of plant responses to soil water deficit in Arabidopsis thaliana permitted the identification of an accession with low sensitivity to soil water deficit

Authors

  • Christine Granier,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Luis Aguirrezabal,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    2. Present address: Facultad de Ciencias Agrarias, Universidad de Mar del Plata – CONICET, Unidad Integrada Balcarce (FCA, UNMP-INTA) cc 276, 7620 Balcarce, Argentina
    Search for more papers by this author
  • Karine Chenu,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Sarah Jane Cookson,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Myriam Dauzat,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Philippe Hamard,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Jean-Jacques Thioux,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Gaëlle Rolland,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Sandrine Bouchier-Combaud,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Anne Lebaudy,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Bertrand Muller,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • Thierry Simonneau,

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author
  • François Tardieu

    1. Laboratoire d’Ecophysiologie des Plantes sous Stress Environnementaux UMR 759, Institut National de la Recherche Agronomique/Ecole Nationale Supérieure d’Agronomie, Place Viala, F−34060 Montpellier, Cedex 1, France;
    Search for more papers by this author

Author for correspondence: Christine Granier Tel: +33 (0)4 996129 50 Fax: +33 (0)4 67522116 Email: granier@ensam.inra.fr

Summary

  • • The high-throughput phenotypic analysis of Arabidopsis thaliana collections requires methodological progress and automation. Methods to impose stable and reproducible soil water deficits are presented and were used to analyse plant responses to water stress.
  • • Several potential complications and methodological difficulties were identified, including the spatial and temporal variability of micrometeorological conditions within a growth chamber, the difference in soil water depletion rates between accessions and the differences in developmental stage of accessions the same time after sowing. Solutions were found.
  • • Nine accessions were grown in four experiments in a rigorously controlled growth-chamber equipped with an automated system to control soil water content and take pictures of individual plants. One accession, An1, was unaffected by water deficit in terms of leaf number, leaf area, root growth and transpiration rate per unit leaf area.
  • • Methods developed here will help identify quantitative trait loci and genes involved in plant tolerance to water deficit.

Introduction

The study of Arabidopsis thaliana genomics is providing new insights into the understanding of the processes involved in plant responses to environmental stimuli. The functional analysis of genes associated with these responses is made possible by the analyses of mutants (Nakazawa et al., 2003) and the natural genetic variation of this species (Koornneef et al., 2004). Methods of high-throughput genetic mapping (Ponce et al., 1999) and large-scale analyses of gene expression (Seki et al., 2002; Kawaguchi et al., 2004) are available. The major challenge is now the phenotypic analysis of this genetic variability, which requires simultaneous analysis of hundreds to thousands of plants.

Phenotyping is relatively easy to do when plant traits are defined in a binary fashion (e.g. flowering/not flowering or dead/alive) and when environmental conditions are defined as the result of a simple action (e.g. light on/off or watered/nonwatered). However, phenotyping is more complex when the traits studied are quantitative (e.g. growth or photosynthesis) and when environmental conditions vary quantitatively or naturally fluctuate (e.g. light intensity or soil water status). The quantitative analysis of complex trait variation requires the establishment of a robust analytic framework, as developed in the field of quantitative genetics (Falconer & McKay, 1996). In this case, the phenotype is not static and better defined by a series of response curves to environmental stimuli (Borel et al., 2001; Reymond et al., 2003; Tardieu, 2003; Hammer et al., 2004). The disadvantage of these ‘response curve’ analyses is that they are time-consuming and require accurate control of environmental conditions so that the only environmental variable changing is the one of interest.

The aim of this paper was to compare response curves of changes in leaf development in response to soil water deficit in a collection of A. thaliana accessions as a ‘proof of concept’. This required methodological and technological advances to accurately impose and quantify the soil water deficit treatments and to establish a robust framework of analysis.

The first problem to be overcome is related to the variation of A. thaliana leaf development in well-watered conditions and the reproducibility of its reduction by soil water deficit, both within an experiment and between experiments. The absolute values defining a phenotype (e.g. leaf area of well-watered plants) depend upon the accurate control of environmental conditions spatially and temporally within a growth-chamber. The effects of water deficit treatment (e.g. the reduction in leaf area) depend not only on the water deficit treatment but also on other environmental conditions affected by the water deficit treatment (for example, water deficit can increase leaf temperature) and this may complicate the analysis.

The second problem to solve concerns the quantification of the stress, which should be defined by physical variables and similar for all studied genotypes. This is not straightforward, especially in the case of soil water deficit. Three methods have been described to impose water deficit on A. thaliana. A first method consists of withholding or limiting irrigation of soil-grown plants (Iuchi et al., 2001; Kawaguchi et al., 2004; Hausmann et al., 2005). Its drawback is that different genotypes do not experience the same degree of water deficit after a few days without irrigation if they deplete the soil water at different rates. An apparently more reproducible method consists of growing plants in a medium whose water potential is lowered with an osmoticum (van der Weele et al., 2000; Kreps et al., 2002). This method, suitable for studying short-term responses, is usually not suitable for longer-term experiments because of the phytotoxicity of the osmoticum and the difficulty of maintaining a constant osmotic potential in such systems with transpiring plants (Verslues et al., 1998). Finally, the method of analysing the responses of detached plants on a bench (Seki et al., 2002) was designed to study short-term changes in gene expression, but is obviously not adapted for studies of plant development. It was shown recently that expression of genes differ in these three systems (Bray, 2004), suggesting that the mode of stress imposition is of crucial importance for analysing the responses mechanisms and for comparing them between genotypes.

The third problem is related to the timing of stress imposition, which should be the same for all genotypes. Sensitivity of leaf expansion or reproductive development to soil water deficit depends on the stage at which the deficit is imposed (Randall & Sinclair, 1988; Ney et al., 1994; Lecoeur et al., 1995; Granier & Tardieu, 1999). In sunflower, final leaf area could be reduced by 10–41% by the same soil water deficit (same intensity and duration) depending on the period of stress imposition (Granier & Tardieu, 1999). It is therefore necessary to impose water deficit at a precise and common stage of development to all genotypes to compare their responses on a robust basis.

Here we describe protocols for the reproducible phenotyping of developmental responses to environmental stimuli in A. thaliana. Methodological concepts were developed and put into practice in an automated system, PHENOPSIS, which is presented and tested. The comparison of response curves of leaf development to water deficit in nine accessions of A. thaliana in a series of independent experiments was done. This analysis allowed the identification of one accession, An-1, which consistently showed a low sensitivity of leaf growth to water deficit in a large range of soil water status. The automated system was also used to analyse other plant traits in response to soil water deficit such as root development and leaf transpiration. An-1 also showed an unusual behaviour in response to soil water deficit in terms of root growth and transpiration rate per unit leaf area.

Materials and Methods

Plant material and growth conditions

Arabidopsis thaliana (L.) Heynh. accessions were grown in a growth-chamber in four experiments (Tables 1 and 2). The Col-0 accession was grown in all four experiments. Cvi-0, Oy-0, Di-M and An-1 were grown in experiments 2, 3 and 4. Mt-0, Ct-1, Bur-0 and Shahdara were grown during experiments 2 and 4 (Table 2). All seeds were provided by M. Simon (Institut National de Recherches Agronomiques, Versailles, France). An-1 and Di-m were selected because they had shown interesting responses to soil water deficit in previous experiments. Other accessions were chosen from a core-collection described by McKhann et al. (2004).

Table 1. Growth conditions during the four experiments performed in the PHENOPSIS system
ExperimentSoil water content (g H2O g−1 dry soil)Daylength (h)Mean air temperature (°C)Daily incident light (mol m−2 d−1)VPDleaf-air (kPa)
  1. For each experiment, mean micrometeorological conditions from sowing to end of rosette leaf development are presented. When different soil water contents were imposed on the plants, they are separated by ‘/’.

10.50/0.40/0.35/0.30/0.25/0.201221.49.10.73
20.45/0.25/0.201218.98.10.80
30.45/0.35/0.30/0.25/0.20/0.151219.38.00.78
40.451220.77.91.05
Table 2. Nine natural accessions used in the four experiments (Experiments 1–4)
Accession no.EcotypeCollection siteLatitude (N)Longitude (E)Experiments
1092Col-0USA52.4 15.11, 2, 3, 4
944An-1Belgium51.1  4.22, 3, 4
919Di-mRussia37.3 55.42, 3, 4
1094Ct-1Italy37.3 15.12, 4
1436Oy-0Norway60.2  6.12, 3, 4
929ShahdaraTajikistan37.3 71.32, 4
1028Bur-0Eire53.1 −9.02, 4
JeaFrance43.4  7.22, 4
902Cvi-0Cape Verde Islands16.0−24.02, 3, 4
1244Ita-0Morocco34.0 −4.12, 4

In all experiments, seeds were sown in cylindrical pots (9 cm high, 4.5 cm diameter) filled with a mixture (1 : 1, v : v) of a loamy soil and organic compost. For the plants subjected to a water deficit treatment, eight 40 × 1-mm slits were made in the sides of the pots facilitating rapid and homogeneous evaporation from the soil during the imposition of treatment. These holed pots were placed in a second intact pot when the desired soil water content was reached (see the Results section).

Light in the growth-chamber was provided by a bank of cool-white fluorescent tubes and sodium lamps and day-length was maintained at 12 h (Table 1). In all experiments, light intensity was measured continuously at plant level, using a light sensor over the waveband of 400–700 nm (LI-190SB; LI COR, Lincoln, NE, USA). Daily incident light intensity was calculated by multiplying the daylength by the mean instantaneous light intensity. Air temperature and relative humidity were measured every 20 s (HMP35A; Vaisala Oy, Helsinki, Finland). Leaf temperature was estimated using copper-constantan thermocouples (0.4 mm diameter) located in the soil before leaf emergence and under the lamina after leaf emergence. All measurements of temperature, light intensity and relative humidity were averaged and stored every 600 s in a datalogger (LTD-CR10 Wiring Panel; Campbell Scientific, Shepshed, UK). Mean leaf-air vapour pressure deficit (VPDleaf-air) was calculated during the light period. Mean micrometeorological conditions are presented in Table 1 for each experiment.

Sowing and plant homogenization

Seeds were imbibed with water before sowing for 1 h and then sown with a pipetman (P1000; Gilson, Middletown, WI, USA) directly onto the soil. From sowing to plant germination, the soil was humidified four times a day with a spray of water. When the second leaf of the rosette had emerged, plants of similar sizes and developmental stages were selected and thinned out to one plant per pot.

Control and measurement of soil and plant water status

Soil water content at retention capacity was measured in a preliminary experiment. Pots were filled with soil, fully wetted and allowed to drain freely. Soil water content was determined by weighing the soil before and after drying (4 d at 180°C). Soil water content at retention capacity was 0.78 g H2O g−1 dry soil and was common to all experiments as composition of soil remained exactly the same. The weight of each pot was measured before and after filling with soil. Soil aliquots were dried to estimate the amount of dry soil and of water in each pot at the time of filling. Subsequent changes in pot weight were attributed to a change in soil water status, after correction for the weight of the plants. For this correction, three plants per accessions and treatments were harvested and weighed weekly. Depending on the experiment soil water content was adjusted to different values on a daily basis (Table 1).

Predawn plant water potential was measured on Col-0 plants at different soil water contents in experiment 1. It was measured when rosette leaf area was fully developed, at the end of the night period while the plant water potential approaches equilibrium with soil water potential. The measurements were done on the inflorescence which was cut and sealed immediately in a pressure chamber (n°3000 Chamber; Soil Moisture Equipment Corp., Santa Barbara, CA, USA). The chamber was pressurized until the sap just wetted the cut surface of the inflorescence. Twenty plants with contrasting values of predawn water potential were selected for abscisic acid (ABA) analysis. Xylem sap was harvested from some of these plants by increasing pressurization in the chamber (at 0.8 MPa above the balancing pressure). Sap samples were stored at −20°C for subsequent ABA analysis. The ABA concentration was analysed in sap samples by radio-immunoassay as described in Quarrie et al. (1988).

Measurements of leaf development

For all accessions, in experiments 1–3, the water deficit treatments started when leaf 6 was initiated on the apex. The number of initiated leaves per plant was counted on five plants per accession, harvested each 2–3 d. The apex was dissected in a drop of water under a stereomicroscope (F8Z; Leica, Wetzlar, Germany) and the number of leaves was counted (excluding the two cotyledons). A leaf was considered to be initiated when it was visible with the microscope at ×160.

From plant germination until the end of vegetative growth, a picture of 12 plants per accession and treatment was taken with a digital camera automatically and once a day. Visible stages of plant development were noted on these pictures according to Boyes et al. (2001). The projected leaf area was measured on the same pictures by an image analysis software (Optimas-Bioscan V. 6–1). The end of rosette leaf development was determined by the cessation of expansion of the last leaf of the rosette.

At the end of their rosette leaf development, the 12 plants were harvested. The leaves were isolated from the rosette, fixed to a piece of paper and scanned. Individual leaf area was measured using the image analysis software and the final rosette leaf area was calculated as the sum of the areas of all individual leaves.

Measurement of root dry weight

In experiment 3, three plants per accession and per water deficit treatment were harvested at the end of their rosette leaf development to determine root dry weight. To facilitate the recovery of a clean root, the root system was carefully washed in water. The roots were then placed in a paper bag and left to dry at 80°C for 4 d. Root dry weight was then measured.

Measurements of transpiration

In experiments 1, 2 and 4, just after bolting, the rosettes of four plants per accession grown at the soil water content of 0.45 g H2O g−1 dry soil were isolated from the soil with a plastic sheet at the end of the dark period. Three hours after the beginning of the light period, pots were weighed every 30 min for 2 h from 10 : 00 h to 12 : 00 h. Plants in the pots were photographed after this period, to measure the total projected leaf area by using the image analysis software. Plant transpiration was calculated by measuring the changes in pot weight as the plastic sheet prevented any changes caused by soil evaporation. Transpiration per unit leaf area (g cm−2 h−1) was then calculated as the ratio between transpiration and projected rosette leaf area, assuming that projected leaf area was a good estimation of transpiring leaf area. In experiments 1 and 4, this was repeated over several days without re-irrigation of the plants, allowing the estimation of transpiration rate per unit leaf area at different soil water contents.

Statistical analysis of data

Each analysis was set with a significance level of P = 0.05 and all statistical analysis was done using the computer package SPSS 11.0 for Windows (SPSS Inc., Chicago, IL, USA). The differences in final rosette leaf area, final leaf area, final leaf number, final root dry weight between the soil water deficit treatments were compared using anova.

PHENOPSIS

The PHENOPSIS automaton is a prototype built by Apilogic (Fondettes, France). It is composed of a steel frame supporting 14 trays with 36 holes (each hole could support a pot, Fig. 1a,b) and a mechanical arm able to move according to a program developed by Apilogic on apigraf ip software. Displacement sensors, a balance (CP622; Sartorius, Goettingen, Germany), a tube for irrigation and a camera (SSC-DC393P; Sony) were loaded onto this arm to weigh, irrigate and take a digital picture of each pot (Fig. 1a–h). Positions of the pots on the 14 trays, dates and times of cycle of irrigation, the weight to be reached by the pot and the necessity to take a picture or not were programmed into a computer on the apigraf ip software (Fig. 1i). Input files for the program could be created in excel or in the apigraf ip software. After each cycle of irrigation, an output EXCEL (Microsoft) file was automatically recorded on the computer containing data such as position of the pot in the chamber, its weight before supplying nutrient solution, the amount of nutrient solution added and the name of the picture file if a digital photograph was taken.

Figure 1.

PHENOPSIS, an automated system for phenotyping Arabidopsis thaliana plants in response to soil water deficit.(a,b) The automated system weighs, waters and takes a digital picture of individual plants. It is composed of a table and an arm which moves according to a program. A camera, a balance and a tube for irrigation are set up on this arm. (c–h) Digital pictures are taken automatically. Here, photographs of the same plant taken at a 6-d interval. (i) Flow-chart representation of the PHENOPSIS program. (j) Spatial distribution of photosynthetic active radiation (PAR) in the PHENOPSIS growth-chamber after correction of heterogeneity with neutral filters. The decrease in light intensity observed on the right of the chamber is caused by the arm of the robot.

The automaton was set up in a growth-chamber built by AM Froid (Montpellier, France). The climatic regulation of the growth-chamber was controlled by a computer with the apigraf ip software. The computer was connected to different sensors such as light, air temperature and air humidity and leaf temperature sensors. Each micrometeorological condition was measured with a 10-s time-lapse and when any climatic variable strayed beyond acceptable limits, as defined in the program, the computer acted to restore the desired environment by employing an air drier or a water sprayer to modify air humidity, an air-cooler or a heater to modify air temperature and the lights to modify day-length. The spatial distribution of the different micrometeorological conditions was measured by displacing the air temperature, light and air humidity sensors from a tray to another one. Gradients of light intensity, air temperature and humidity were corrected by adding neutral filters and by fine-tuning air circulation and orientation of the sprays in the chamber (shown for light intensity in Fig. 1j).

Results and Discussion

Growing plants with reproducible final leaf size at different soil water contents

Final leaf size in Arabidopsis thaliana is affected by micrometeorological conditions such as incident light (Cookson et al., 2005). Temperature also affects the rate of leaf development (Granier et al., 2002). Heterogeneity of micrometeorological conditions within a growth chamber and differences in these conditions among experiments can cause variation in leaf development of well-watered plants but also differences in the response of plants to soil water deficit. For example in maize, a reduction in leaf expansion rate by soil water deficit is more drastic at high VPD (Ben Haj Salah & Tardieu, 1997). The spatial and temporal control of micrometeorological conditions is therefore a requirement for obtaining reproducible phenotypes.

In the PHENOPSIS growth-chamber, light intensity, temperature and humidity were homogenized and could be reproduced from an experiment to another one. As a consequence, when Col-0 plants were grown in experiments 1, 2 and 3 in the PHENOPSIS system, their leaf development in well-watered conditions and their responses of leaf development to soil water deficit were similar (Fig. 2). Final leaf 6 area was not significantly affected by soil water contents between 0.50 and 0.30 g H2O g−1 dry soil but was consistently reduced below this threshold.

Figure 2.

Independent experiments, experiment 1 (diamonds), experiment 2 (upward-pointing triangles) and experiment 3 (downward-pointing triangles) illustrating reproducibility of response curves. Relationship between final leaf 6 area and soil water content for Col-0 plants grown in three independent experiments; 95% confidence intervals of are presented (n = 12).

Imposing rapid and well-defined soil water deficits to different genotypes

The difficulty of analysing the responses of different genotypes to soil water deficit when water stress experiments are based on withholding irrigation from soil-grown plants is shown in Fig. 3a. Soil water depletion rate was measured at the same developmental stage of plant development in the nine A thaliana accessions. The depletion rate of An-1 plants was sixfold lower than that of Bur-0 plants at stage 3.90 (fully developed rosette, as described in Boyes et al., 2001). This was due to differences in total plant transpiration rate, which was associated with differences in leaf area (Fig. 3b) and to a lesser extent with differences in transpiration rate per unit leaf area (Fig. 3c). After 4 d without irrigation, An-1 and Bur-0 experienced soil water contents of 0.4 and 0.2 g H2O g−1 dry soil, respectively (Fig. 3a, inset), thus each accession experienced very different water deficits. This could considerably bias the comparison of sensitivities of different genotypes to water stress (Ray et al., 1997; Lacape et al., 1998).

Figure 3.

Joint analysis of soil water depletion, leaf area and transpiration per unit leaf area in the nine accessions of Arabidopsis thaliana grown in experiment 2 at a soil water content of 0.45 g H2O g−1 dry soil. (a) Soil water depletion rate estimated during the transpiration assay in experiment 2 (see the Materials and Methods section). Inset, Changes with time in soil water content (SWC) after withholding irrigation at stage 5.10 (first flower buds visible) for four accessions (An-1, downward-pointing triangles; Col-0, upward-pointing triangles; Shahdara, circles; Bur-0, squares) during experiment 2. (b,c) Projected leaf area and transpiration rate per unit leaf area. For each accession, all measurements were done at stage 5.10 (first flower buds visible, according to Boyes et al., 2001). Data are given as means with 95% confidence intervals (n = 4).

The PHENOPSIS system allows the imposition of water deficits over the whole cycle of soil-grown plant development and the production of stable and reproducible water potentials to roots. The principle is to control and stabilize the soil water status in a large number of transpiring plants of different genotypes (see also the Material and Methods section for details). As in other studies, individual pots were weighed at regular intervals and pots were watered with the amount necessary to reach a given soil water content (Granier & Tardieu, 1999; Yu & Setter, 2003).

The daily water delivery was managed independently for each pot so plants with high transpiration rate received more water than those with a low transpiration rate. This protocol could be performed manually on a small scale but has been automated in the PHENOPSIS system with the aim of simultaneously studying more than 500 plants. It took approx. 1.5 h for the automaton to weigh, water and take a digital image of 500 pots (vs 7 h for 200 plants using a manual procedure). Nine accessions with contrasting leaf development and stomatal conductances (the same accessions as those presented in Fig. 3) were grown in this system. Despite differences in leaf area and transpiration rates, rigorously identical and stable soil water contents (0.45, 0.25 and 0.20 g H2O g−1 dry soil) were imposed on the different accessions over a long period (i.e. for 30–50 d, Fig. 4). The daily oscillation in soil water content was less than 0.05 g H2O g−1 dry soil in all accessions at the beginning of the experiment but increased towards the end of the experiment.

Figure 4.

Nine accessions were grown at three different and stable soil water contents identically imposed on each accession. Changes with time in soil water content imposed on the nine accessions in experiment 2. Plants were grown in well-watered conditions (soil water content adjusted daily to 0.45 g H2O g−1 dry soil) and two water deficit treatments (soil water content adjusted daily to 0.25 or 0.20 g H2O g−1 dry soil). Periods of stress started at leaf 6 initiation for all accessions. Data are given as means with 95% confidence intervals (n = 12).

As explained in the Material and Methods section, the protocol used to impose soil water deficit required that some plants were killed during the experiments to take into account plant weight in the calculation of the amount of nutrient solution to add to a given pot. In the different experiments presented here, this was done weekly on three plants per accession and per treatment. At the end of each experiment, the calculated soil water content was verified in each pot by weighing wet and dry soil after the plant was removed (Fig. 5).

Figure 5.

Comparison of estimated and measured soil water contents.

Physiological indicators of water stress

A methodological alternative to quantify soil water deficit could have been to measure physiological indicators of stress rather than the soil water content, and to drive watering requirements by these indicators. For example, measurement of leaf temperature could be considered as a surrogate of measurements of stomatal conductance (Merlot et al., 2002). The rapid progress of in situ imaging will probably allow one to noninvasively evaluate other physiological characteristics of a large number of plants in the future. However, there is a theoretical problem, namely that physiological measurements themselves depend on the whole-plant controlling systems. For example, stomatal conductance may or may not be an indicator of water deficit depending on the species or genotype considered (Tardieu & Simonneau, 1998). Similarly, a high concentration of ABA in the xylem sap ([ABA]xyl) can indicate that the studied genotype is currently under stress or that it has a high constitutive biosynthesis of ABA (Tardieu, 2003). Each physiological variable has its own response to soil water status (e.g. stomatal conductance, photosynthesis, plant water potential, osmotic adjustment (Hsiao, 1973) and organs of the same plant may display different responses for the same physiological variable (e.g. osmotic adjustment, Westgate & Boyer, 1985). This casts doubt on the value of choosing one particular physiological variable as a stress indicator with a solid theoretical base. We therefore propose that the driving variables for stress imposition should be the conditions imposed at the boundary of the plant (such as soil water potential or evaporative demand) and therefore that they should be similar in all genotypes studied, as is the case in the PHENOPSIS system. In this way, physiological variables would be evaluated together with the phenotypic variables of interest. For example, the concentration of ABA in the xylem sap and leaf transpiration were estimated in Col-0 plants grown at different soil water contents and were related to the soil water potential (Fig. 6). Soil water content, which is the controlled variable in the PHENOPSIS protocol can be transformed into soil water potential afterwards by the water-release curve obtained for the given substrate (Fig. 6a).

Figure 6.

Interpretation of soil water content. (a) Water release curve: relationship between predawn water potential and soil water content. Each point is the value measured on an individual plant at the end of experiment 1. Soil water content was measured by weighing the soil before and after drying immediately after the plant was removed. (b) Relationship between concentration of abscisic acid (ABA) in the xylem sap and soil water content. Each point represents the value measured on an individual plant at the end of experiment 1. (c) Relationship between transpiration rate per unit leaf area and soil water content. Data are given as mean values with 95% confidence intervals (n = 4).

Timing of stimuli imposition and phenological stage of genotypes

The nine accessions presented in Fig. 3 had different developmental rates (Fig. 7). The time taken to reach leaf 6 initiation (Fig. 7a), leaf 6 emergence (stage 1.06, as defined by Boyes et al., 2001, Fig. 7b) or emergence of the first flower buds (stage 5.10, as defined by Boyes et al., 2001, Fig. 7c) differed among them and their developmental rates between each of these stages also differed. In a series of experiments on A. thaliana and sunflower, the only way to get a robust model of leaf growth response to soil water deficit or temperature in different experiments was to express the timing of stress in time after leaf initiation (Granier & Tardieu, 1998, 1999; Granier et al., 2002). This stage of leaf development is not easy to identify as it needs destructive measurements. However, it is more appropriate than the visible stages defined by Boyes et al. (2001) which do not necessarily correspond to a common leaf age in all accessions. For example, it is clear from Fig. 7 than when leaf 6 emerges in Cvi-0, this leaf has spent more time hidden than in other accessions (Fig. 7a,b). To avoid this problem, water deficit treatments were imposed at leaf 6 initiation in all experiments and for all accessions. It corresponded to different calendar dates depending on the accession. For example, it occurred 9 d after plant emergence for Cvi-0 and 12 d after plant emergence in Bur-0 in Experiment 2 (Fig. 7a).

Figure 7.

Nine accessions grown in well-watered conditions have different developmental rates and reached common stages of development (according to Boyes et al., 2001) at different dates. (a) Time to initiate leaf 6. (b) Time to reach stage 1.06, corresponding to ‘6th leaf < 1mm’. (c) Time to reach stage 5.10 corresponding to ‘first flower buds visible’. In (b) and (c), data are given as mean values with 95% intervals of confidence (n = 12).

An example of the analysis of the genetic variability of responses to soil water deficit: identification of An-1 as a potentially drought-tolerant accession.

In Experiment 2, nine accessions were grown at a constant soil water content of 0.45 g H2O g−1 dry soil from plant emergence to leaf 6 initiation. When leaf 6 was initiated, plants were subjected to three different soil water contents: 0.20, 0.25 and 0.45 g H2O g−1 dry soil until the end of rosette development (Fig. 4). Rosette leaf area was reduced with declining soil water status in eight of the nine accessions (Fig. 8). It was slightly but not significantly reduced by the moderate water deficit treatment in Sha and Ct-1 and was significantly increased in An-1. For the most severe water deficit treatment, rosette leaf area was significantly reduced in eight accessions with considerable variability in this response. It was reduced by a factor of 10 in Bur-0 and only by a factor of 2 in Oy-0 and Col-0. Conversely, the final leaf area of An-1 was unaffected or even increased for the same water deficit treatment (Fig. 8).

Figure 8.

Final leaf area of the rosette in the nine accessions subjected to three soil water contents (experiment 2). (a) Pictures of the accessions grown in the PHENOPSIS system subjected to soil water contents of 0.45 (left), 0.25 (middle) and 0.20 (right) g H2O g−1 dry soil. Three plants are shown for each accession in each treatment at the end of their rosette development. The inflorescence was removed to clearly see the rosette. (b) Final leaf area (sum of all individual leaf areas) of the accessions grown in experiment 2 at three soil water contents. Data are given as mean values with 95% confidence intervals (n = 12).

An additional experiment (experiment 3) was performed to check the robustness of the method, to enlarge the range of soil water deficits and to enlarge the number of plant traits measured in response to water deficit. In Experiment 3, plants of five accessions were subjected to six stable soil water contents when leaf 6 was initiated. As shown for Col-0 (Fig. 2), final rosette leaf area was not affected in the range of soil water contents from 0.45 to 0.30 g H2O g−1 dry soil in all five accessions. For An-1, leaf area was actually increased for soil water contents of 0.25 g and 0.20 g H2O g−1 dry soil as in experiment 2 (Figs 9 and 8). It was reduced by a factor of 2 in the most severe water deficit (0.15 g g−1 g H2O g−1 dry soil, corresponding to −1.0 MPa), while leaf area of other accessions was reduced by a factor of 8 at the same soil water status. A two-way anova was done with the factors ‘accession’ and ‘soil moisture’ on the data from experiments 2 and 3. Because An-1 was significantly different from the others the main effects of the accession (F4,145 = 187.4, P > 0.0005) and soil moisture (F2,147 = 73.4, P > 0.0005) were significant as was the interaction between them (F8,135 = 13.0, P > 0.0005). These values are reported for experiment 2 but were similar for experiment 3.

Figure 9.

Plant responses to soil water deficit differed among five accessions and allowed the identification of an accession with unusual phenotype (An-1). (a) Relationship between final rosette leaf area (= sum of individual leaf areas) and soil water content (experiment 3). (b) Relationship between final leaf number per rosette and soil water content (experiment 3). (c) Relationship between root dry weight and soil water content (experiment 3). (d) Relationship between transpiration rate per unit leaf area and soil water content (experiment 4). For each accession, transpiration rate per unit leaf area is expressed in percentage of its value at a soil water content of 0.45 g H2O g−1 dry soil (TRmax). The values of TRmax are given in the inset. In the four plots, each symbol represents an accession: downward-pointing triangles, An-1; circles, Col-0; diamonds, Oy-0, squares, Cvi-0; upward-pointing triangles, Di-m. In (a), (b) and (c), data are means with 95% confidence intervals (n = 12 in a and b; n = 3 in c); in (d), each point represents an individual plant.

The leaf number per plant was not affected in a range of soil water contents from 0.45 to 0.25 g H2O g−1 dry soil in all five accessions. It was significantly reduced in four of the five accessions when they were grown at grown at 0.15 g H2O g−1 dry soil. At this very low soil water content, it was reduced by a factor 1.3 (Di-m) to 1.7 (Oy-0). An-1 did not reduced its total leaf number even in very severe water deficit conditions. The total root dry weight was significantly increased by moderate soil water deficit in Oy-0, Di-m and An-1 but was not significantly affected in Col-0 and Cvi-0. Stimulation of primary root elongation and total root dry weight by low water potential in nutrient-agar medium has also been reported by van der Weele et al. (2000) in A. thaliana. In our case, maximal root dry weight was reached at different soil water contents depending on the accession. It was close to 0.45 g H2O g−1 dry soil for Col-0 but close to 0.30 for Oy-0 and 0.20 for An-1. Root development in An-1 was therefore stimulated by very low soil water contents compared with the other accessions. Finally, the behaviour of An-1 also differed from the other accessions in terms of transpiration per unit leaf area under water deficit (Experiment 4, Fig. 9d). For all accessions studied the transpiration rate per unit leaf area was maximal when plants were grown at 0.45 g H2O g−1 dry soil. These maximal values differed among accessions (Fig. 9d, inset). Transpiration rate per unit leaf area was significantly decreased in Oy-0, Col-0, Cvi-0 and Di-m for soil water contents below 0.35 g H2O g−1 dry soil. By contrast, it remained at its maximal value in An-1 plants for soil water contents between 0.45 and 0.25 g H2O g−1 dry soil.

Had the experiments done here been based upon the addition of constant amounts of water to impose soil water deficit to the different genotypes, the low transpiration rate of An-1 would have resulted in wetter soils for this accession than the others (Fig. 3). Thus, its lack of response to soil water deficit could have been an artefact of the low rate of water use of this accession. However, this is not the case here, as in the PHENOPSIS system conditions were maintained equivalent for the different genotypes at the plant boundary. The response of An-1 is not only a result of its low constitutive leaf area; another accession with low constitutive leaf area, Ct-1, reduced its final leaf area by a factor of 2 for a soil water deficit treatment that increased leaf area by a factor 2 in An-1 (Fig. 8).

Conclusions

PHENOPSIS was used to dissect plant responses to soil water deficit in a collection of natural accessions of A. thaliana. Stability of soil water content and micrometeorological conditions during the experiments combined with the imposition of the water deficit periods at a precise stage of plant development was necessary for reproducibility of the response curves. For all the traits measured in our analysis, namely leaf growth, leaf number, transpiration rate per unit leaf area and root growth, An-1 had a particular behaviour. It cannot be stated at this stage whether the low sensitivities of An-1 to water deficit for the four traits are caused by a cascade of physiological events (all responses having causal relationships between them), an evolutionary process (natural selection in this accession for alleles of tolerance to water deficit in these four traits) or by a common genetic determinism of the four traits by the fixation of pleiotropic alleles (as suggested by McKay et al., 2003 for the correlated evolution of δ13C and flowering time). An analysis involving mapping populations and detection of quantitative trait loci could test each of these three hypotheses. Recent papers suggest that quantitative genetics is feasible for analysing growth-related traits in A. thaliana (El-Lithy et al. 2004), and their interaction with irrigation treatment (Hausmann et al., 2005).

An obvious drawback of the method presented here is that, to ensure reproducibility of phenotype measurements, all environmental conditions except soil water status were kept to a standard value. The results obtained in this way may therefore seem difficult to extrapolate to other, more complex, situations. A solution of this problem may be to combine several response curves in order to predict the combined effect of several environmental stimuli in a modelling approach. This was done for maize leaf growth in which the effects of temperature, VPD, light intensity and soil water status were dissected and then combined in a model (Tardieu et al., 2000; Reymond et al., 2003). After wide-ranging response curve characterization, such predictive models could be tested with the PHENOPSIS system by comparing the performance of different genotypes under programmed fluctuating scenarios simulating those in natural conditions.

Acknowledgements

This research was supported by grants from the Institut National de Recherches Agronomiques (INRA, ‘Ecogene’), GABI-GENOPLANTE (AF 2001 094) and the European Research Training Network (HPRN-CT-2002-00267). L. A., S. B. and S. C. were supported by grants from the Institut National de Recherches Agronomiques (département Environnement et Agronomie), from GENOPLANTE and from the European Commission, respectively.

Ancillary